Fix alternate schema bug in sql proxy #296
Conversation
| var rs = metadata.getPrimaryKeys(null, schemaName, tableName); | ||
| if (!rs.next()) | ||
| throw new SQLException("No primary key found for " + table); | ||
| throw new SQLException("No primary key found for " + (schemaName != null ? schemaName + "." + tableName : tableName)); |
There was a problem hiding this comment.
Should we make this into a function schemaName != null ? schemaName + "." + tableName : tableName ? Since it is repeated a couple of times.
|
@danielgerlag are there no changes required to the releation reactivator to distinguish between tables with the same name but from different schema? |
| throw new SQLException("No primary key found for " + (schemaName != null ? schemaName + "." + tableName : tableName)); | ||
| var mapping = new NodeMapping(); | ||
| mapping.tableName = rs.getString("TABLE_SCHEM") + "." + tableName; | ||
| String actualSchema = rs.getString("TABLE_SCHEM"); |
There was a problem hiding this comment.
Is TABLE_SCHEM a typo that we are perpetuating?
There was a problem hiding this comment.
TABLE_SCHEM is part of the JDBC drivers, for returning metadata.
https://learn.microsoft.com/en-us/sql/connect/jdbc/reference/getschemas-method?view=sql-server-ver17#remarks
The reactivator does not have this issue, because we are not manually constructing a query string, the debezium library handles things. |
3 participants
This pull request refactors the way tables and schemas are handled in the SQL proxy by introducing a new
TableRegistryclass and updating the table cursor logic to support schema-qualified table names. The changes improve robustness when dealing with tables from multiple schemas and ensure that only registered tables are processed, and fixes a bug where tables in the non-default schema could not be bootstrapped.Schema and Table Registry Improvements:
TableRegistryclass to manage table-to-schema mappings, parsing thetablesconfig and providing lookup methods for table existence and schema name. (sources/relational/sql-proxy/src/main/java/io/drasi/TableRegistry.java)ResultStreamto instantiate and useTableRegistry, checking if each requested table exists and retrieving its schema before creatingTableCursorinstances. Unregistered tables are logged and skipped. (sources/relational/sql-proxy/src/main/java/io/drasi/ResultStream.java) [1] [2]Table Cursor and Query Handling:
TableCursorto accept bothschemaNameandtableNamein its constructor, and to use these values when building SQL queries and reading table mappings. (sources/relational/sql-proxy/src/main/java/io/drasi/TableCursor.java)TableCursorto use fully qualified table names (including schema when present), ensuring correct querying in multi-schema environments. (sources/relational/sql-proxy/src/main/java/io/drasi/TableCursor.java)sources/relational/sql-proxy/src/main/java/io/drasi/TableCursor.java)